Line Drawing, Leap Years, and Euclid

MITCHELL A. HARRIS
Technical University of Dresden

AND

EDWARD M. REINGOLD
Illinois Institute of Technology

Abstract. Bresenham’s algorithm minimizes error in drawing lines on integer grid

points; leap year calculations, surprisingly, are a generalization. We compare the two
calculations, explicate the pattern, and discuss the connection of the leap year/line
pattern with integer division and Euclid’s algorithm for computing the greatest
common divisor.

Categories and Subject Descriptors: F.2.1 [Analysis of Algorithms and Problem

Complexity]: Numerical Algorithms and Problems—Number-theoretic computations
(e.g., factoring, primality testing); I.3.3 [Computer Graphics]: Picture/Image
Generation; J.2 [Physical Sciences and Engineering]: Astronomy; Mathematics and
statistics
General Terms: Algorithms, Theory
Additional Key Words and Phrases: Bresenham’s algorithm, calendar algorithms,
continued fractions, Euclid’s algorithm, greatest common divisor, leap years, line
drawing, scan-line conversion

1. INTRODUCTION that is discrete with as little error as pos-

Bresenham’s algorithm [Bresenham 1965]
is the classic technique for plotting lines on
bitmaps: it approximates linear segments
defined by rational coefficients using only
integral points. The algorithm calculates
a finite set of points on the integer lat-
tice with minimum total vertical distance
from the original line segment. The essen-
tial design features of the algorithm are
minimization of error (it converts a line
segment from a continuous domain to one
sible), speed (it uses only integer arith-
metic), and parsimony (as few pixels as
possible are used to ensure adjacency).
In a very different domain, leap year cal-
culations [Reingold and Dershowitz 2001]
involve a similar mapping from the con-
tinuous to the discrete. By astronomical
observation, one can calculate the ratio
between the duration of the revolution
of the earth around the sun (the year),
and the duration of the rotation of the
earth (the day); there are approximately

Authors’ addresses: M. A. Harris, Department of Computer Science, Technical University of Dresden, D-

01062 Dresden, Germany; email: harris@tcs.inf.tu-dresden.de; E. M. Reingold, Department of Computer
Science, Illinois Institute of Technology, Stuart Building, 10 West 31st Street, Suite 236, Chicago, IL 60616-
2987; email: reingold@iit.edu.
Permission to make digital/hard copy of part or all of this work for personal or classroom use is granted
without fee provided that the copies are not made or distributed for profit or commercial advantage, the
copyright notice, the title of the publication, and its date appear, and notice is given that copying is by
permission of ACM, Inc. To copy otherwise, to republish, to post on servers, or to redistribute to lists requires
prior specific permission and/or a fee.
°2004
c ACM 0360-0300/04/0300-0068 $5.00

ACM Computing Surveys, Vol. 36, No. 1, March 2004, pp. 68–80.
Line Drawing, Leap Years, and Euclid
365.25 days per year. But calendars, by de-
sign, require an integral number of days
per year; to account for the extra quar-
ter of a day, we occasionally add an ex-
tra day to the year. Arithmetical calen-
dars such as the Julian and the Jewish,
and the common arithmetical approxima-
tion to the Islamic calendar, have special
formulæ for specifying exactly which years
get extra days; these calendars add extra
days as evenly as possible over the course
of years—we will see what “evenly” means
in the discussion of calendars in Section 2.
How are the line segment and the leap
year calculations related? It turns out that
Bresenham’s algorithm computes a spe-
cial case of leap year calculations. They
are both computing approximations to
a line with rational slope using integer
division. Both the repeating pattern of
dots in a rasterized line segment and
the repeating pattern of leap years can
be computed using integer division—and
they are intimately related to Euclid’s
algorithm for calculating greatest com-
mon divisors and continued fractions. Not
surprisingly, others have noticed simi-
larities among these three areas. The
connection between astronomical cycles
and greatest common divisors is folkloric
[Rockett and Szüsz 1992], going even back
to the Greeks. More recently, Brons [1974],
Castle and Pitteway [1987], Pitteway
[1985], and Troesch [1998] all discussed
the connection between line drawing and
Euclid’s algorithm. Here we give explicit
correspondences among all three. First,
we discuss the details of each domain, and
then compare them.
2. LINE DRAWING
In this section we derive Bresenham’s line
drawing algorithm, one of the fundamen-
tal algorithms of computer graphics. In its
final form, as usually presented, the algo-
rithm is obscure; deriving it in a step-by-
step way is both interesting and illuminat-
ing. Furthermore, it is much easier to deal
with the simpler initial form than with the
equivalent final form.
A line drawing algorithm has as its in-
put the integer starting point (xS , y S ) and
ACM Computing Surveys, Vol. 36, No. 1, March 2004.
69
the integer ending point (xE , y E ) of a line
segment. The basic algorithm calculates
the slope of the line,
1y yE − yS
= .
1x xE − xS
For simplicity we assume that 0 < 1 y ≤
1x 6= 0: if either 1x = 0 or 1 y = 0, the
line drawing is trivial; if 1 y > 1x, we
merely interchange the x and y axes. To
draw the line segment, we want to darken
the xE − xS + 1 pixels specified by
µ ¶
1y
y = y S + round (x − xS ) ,
1x
x = xS , xS + 1, . . . , xE . (1)
Traditionally, this sequence of pixels is
encoded as a sequence of the plotter con-
trols “east” and “northeast.” We prefer to
separate the changes of x and y into just
“east” and “north” movements.
We start with (x, y) = (xS , y S ), and cal-
culate the sequence of pixels incremen-
tally, repeatedly adding 1 to x, and oc-
casionally adding 1 to y. We make the
decision to increment y by keeping track
of the error ², the vertical distance of the
pixel to the true line segment. If ² ≥ 1/2,
the pixel above more closely approximates
the true value of y, so we increment y by
1 and decrement ² by 1. Otherwise, we add
the slope 1 y/1x (which is between 0 and
1 by assumption) to ² and do not change
y. The initial error is ² = 0 and this al-
gorithm minimizes the total vertical error
of the pixels to the real line, since induc-
tively the error ² is always in the range
[−1/2, 1/2], the best possible on the in-
teger lattice. This algorithm is shown in
Algorithm 1(a); it computes the points
given by Equation (1).
Algorithm 1(a) can easily be restricted
to integer operations—calculate the (ra-
tional) slope 1x/1 y in lowest common
terms, so that 1x and 1 y are relatively
prime, and multiply all assignments and
tests involving ² by 21x; for convenience
we also shift the range of ² from [−1x, 1x]
to [0, 21x]; the result is Algorithm 1(b).
70
Algorigthm 1. Transformations leading to Bresenham’s algorithm. Version (a) uses floating point operations,
eliminated in version (b) by multiplying through by 21x and shifting the resulting range of ² from [−1x, 1x)
to [0, 21x); version (c) results from then moving the update of the error to after the comparison, shifting the
initial value of ², and using a zero test in the if.
This algorithm can be converted, in turn,
to the traditional Bresenham algorithm of
Algorithm 1(c) by moving the update of the
error after the comparison, shifting the
initial value of ², and using a “zero test”
(this simpler test is the point of the trans-
formation). Each of these transformations
preserves correctness.
Examining Algorithm 1(c), we see that
y is incremented by 1 only when ² is decre-
mented by 21x, making y the quotient on
division by 21x, with remainder ². Hence,
the equation plotted by the algorithm cor-
responds to Equation (1):
¹ º
2n1 y + ²0
yn = yS +
21x
¹ º
2(xn − xS )1 y + 1x
= yS + , (2)
21x
Since at every step an increment is made
to x, and possibly to y, the resulting line
drawing is “connected” in that successive
points are adjacent horizontally or diag-
onally. As an example, three lines with
different right endpoints are shown in
Figure 1. Note the simple pattern for the
line from (0, 0) to (30, 10) in Figure 1(a)
and the more complicated patterns for
lines from (0, 0) to (30, 11) and (30, 12) in
Figures 1(b) and 1(c), respectively.
We call a set of points generated by
Bresenham’s algorithm a Bresenham line.
Harris and Reingold
Because of the manner in which x and y
are incremented, every x coordinate and
every y coordinate between the endpoints
occurs in a Bresenham line. All such lines
are made of horizontal segments of length
one or more; the lengths of the segments
in a particular Bresenham line have no
more than two distinct values, excluding
the segments touching the endpoints. If
1 y = 1, then the line has a uniform ap-
pearance, with a single length for the seg-
ments. For 1 y > 1, the pattern of alter-
nating long and short sequences appears
to have some pattern but it is difficult
to determine by inspection. Changing the
initial value of ² does not change the basic
pattern, but shifts it.
Bresenham’s original paper [1965] es-
sentially gives Algorithm 1(c) directly
from an analysis of error. Sproull [1982]
derived Bresenham’s algorithm as we do,
starting with the straightforward naı̈ve
algorithm, using rationals and trans-
formed it step by step into Bresenham’s
by scaling and shifting. Berstel [1990]
described the Bresenham patterns us-
ing formal language theory; these pat-
terns are an instance of Sturmian
words—see Allouche and Shallit [2003],
Section 9.2.
3. LEAP YEARS
Humans observed millennia ago that
there are approximately 365.25 days
ACM Computing Surveys, Vol. 36, No. 1, March 2004.
Line Drawing, Leap Years, and Euclid 71

Fig. 1. Three examples of lines and their approximation by points using Bresenham’s algo-
rithm; the circled lattice points in a plot constitute the “Bresenham line” approximating the
dashed line shown.

between successive winter solstices; hence
there are nearly 1461 days for every
4 years. Since 1461 − 365 × 4 = 1, it
takes 4 years for the extra quarter of a
day to add up to a full day, so an extra
day must be added every four years. This
corresponds to the (old style) Julian cal-
endar leap year rule: year y is a leap
year if and only if y ≡ 0 (mod 4), and
to the Coptic/Ethiopic rule that year y
is a leap year if and only if y ≡ 3
(mod 4). The given ratio of 1461 days
every 4 years gives an average year
length of 365.25 days per year, achieved
with an integral number of days per
year.

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

72
But an average of 365.25 days per year
is not fully accurate. The true average
(which is changing very slowly over time)
is closer to 365.242 days per year and
that small difference accumulates into an
error of about one day every 125 years.
By the sixteenth century the error had
caused the spring equinox to shift from
its traditional (Roman) date of March 21
to around March 11; if uncorrected, this,
in turn, would cause the date of Easter
(which depends on the equinox) to migrate
through the seasons. Pope Gregory XIII
reset the calendar by 10 days, refined
the definition of the date of Easter, and
changed the leap year rule to omit leap
days in century years not divisible by
400. A more thorough discussion can
be found in Reingold and Dershowitz
[2001], which also gives extensive
references.
The leap year rule for the Gregorian cal-
endar (our present calendar) is that leap
years are those divisible by 4, except not
those divisible by 100, except those divisi-
ble by 400. This rule gives an average year
length of 365 + 1/4 − 1/100 + 1/400 =
36597/400 = 365.2425 days. See Shallit
[1994] for a discussion of the mathemat-
ics of Gregorian-like leap year rules. The
Gregorian rule more closely approximates
the astronomical average in the long run,
but with a more significant deviation from
that average (see below).
Other calendars use different ratios and
so have different leap year rules. In the
Islamic calendar, a year is defined as the
passage of twelve lunations (times be-
tween successive new moons) [Reingold
and Dershowitz 2001]. This results in ap-
proximately 354 11 days per year or about
30
10631 days every 30 years. Since 10631 −
354 × 30 = 11, we need 11 leap years in
every 30-year cycle. The common arith-
metical approximation to the Islamic cal-
endar has leap years in the 2nd, 5th, 7th,
10th, 13th, 16th, 18th, 21st, 24th, 26th,
and 29th years of the cycle. There is an
arithmetical rule for this selection but it
is not obvious: the year y is a leap year
if and only if ((11 y + 14) mod 30) < 11.
Though this seems arbitrary, the formula
follows from the ratio of years to days.
Harris and Reingold
In our presentation below, we use this
arithmetical approximation to the Islamic
calendar as the canonical example since
it displays more of the depth and gen-
erality of the derivation than does the
Julian.
The task of a calendar leap year rule is
to spread the leap years as evenly as possi-
ble over the cycle of years. We assume that
the physical parameters never change and
that there is at least one day per year.
Given a ratio of x days in c years, where
x and c are positive integers, c ≤ x, we
seek an even distribution of leap years.
The length in days of a normal year is
L = bx/cc, and a leap year has L + 1
days. If c divides x evenly, no leap years
are needed; otherwise,
x = cL + l ,
where 0 < l < c are integers; l is the num-
ber of leap years that must occur in a cycle
every c years starting with year 0. So, we
insist that a year be a leap year if its num-
ber has reached or just passed an integral
multiple of the ratio c/l ; that is, year y is
the kth leap year (with year 0 as the first
leap year) if
c c
(k − 1) ≤ y < (k − 1) + 1, (3)
l l
that is, if
(k − 1)c ≤ ly < (k − 1)c + l
or
ly mod c < l . (4)
We can also generalize this by insisting
that year 0 be in position s of the leap year
pattern; we shift:
l ( y + s) mod c < l . (5)
In the arithmetical approximation to the
Islamic calendar,
11( y + 4) mod 30 < 11,
which is inequality (5) with s = 4. In-
equality (5) also works for the cycle of
ACM Computing Surveys, Vol. 36, No. 1, March 2004.
Line Drawing, Leap Years, and Euclid
Hebrew months and years, as well as
Julian, Coptic, and Ethiopic leap years
(see Reingold and Dershowitz [2001]). It
does not work for the Gregorian leap years,
though it does apply to the positions of
long (31-day) months versus short months
on the Gregorian calendar—see Reingold
and Dershowitz [2001].
The notion of “even distribution” here
should be contrasted with the division
into “as-equal-as-possible parts” given in
Graham et al. [1994], Equation 3.24, to di-
vide lines of text into balanced columns.
Translating their division into the leap
year setting, the cL + l days would be di-
vided into c years as
» ¼ » ¼ ample, the number of days in years 0, 1,
cL + l cL + l − 1
cL + l = + + ···
c c
» ¼
cL + l − c + 1
+ ,
c 11 × 5 + ((11 × 3) mod 30)
putting all the leap years at the start of the
cycle. On the other hand, comparing the
leap year placement of the Gregorian cal-
endar with that of the Julian calendar—
inequality (5) with l = 1, c = 4, s =
0—we find the year-by-year error in the
Julian calendar cycles through the val-
ues 0, 1/4, 1/2, 3/4. The corresponding
cycle of errors for the Gregorian calen-
dar is 0, 97/400, 2 × 97/400, 3 × 97/400,
4 × 97/400 − 1 = −3/100, . . . ; these er-
rors range from −18/25 at year 96 to
591/400 at year 303. The Gregorian wob-
ble around its average is thus 591/400 +
18/25 = 879/400, almost 2.2 days com-
pared to Julian calendar’s wobble of 0.75
days around its average. We want the leap
years distributed as uniformly through-
out the cycle so the range of errors is
minimized.
So far we have just considered years.
We also want to find relationships between
days and years. Just as we label the years
0, 1, 2, . . . , so we label days 0, 1, 2, . . . ,
with day 0 being the first day of year 0.
For example, how many days are there
in the range of years 0 · · · y? To compute
this, we need to know the number of leap
years k in the range of years 0 · · · y. For
the case s = 0, inequality (3) can be rear-
ACM Computing Surveys, Vol. 36, No. 1, March 2004.
73
ranged to give
¹ º
y
k= + 1. (6)
c/l
Using this formula, we find the number of
days x in the years in the range [0 · · · y] is
x = L( y + 1)
¹ º
l ( y + 1) + ((l (s − 1)) mod c)
+ ; (7)
c
this sums L days for each of the y +1 years
and an extra day for each leap year (see
Reingold and Dershowitz [2001]). For ex-
2, 3, 4 (or the number of the first day of
year 5) in the arithmetical approximation
to the Islamic calendar (with s = 4) is
¹ º
354 × 5 +
30
= 1770 + b58/30c = 1771,
where year 2 is a leap year and both years
and days are numbered starting from 0.
We can also go the other direction, find-
ing the year at a particular day num-
ber. A derivation similar to the above
is followed; the details can be found in
Reingold and Dershowitz [2001], Sec-
tion 1.12. We find that
y = yS
¹ º
c(x − xS ) − ((ls − l ) mod c) + c − 1
+ ,
cL + l
(8)
where xS is the day number we specify
must begin year number y S . Contrary to
the convention used here, Reingold and
Dershowitz [2001] require that the first
day of year 1 (as opposed to year 0) be
numbered day 0; this makes our functions
shifted versions of theirs. For example,
in Equation (8), we achieve this by let-
ting y S = 1 and xS = L (our day num-
ber for the first day of our year 1) which
yields Equation (1.64) from Reingold and
Dershowitz [2001].
74
Fig. 2. Points of the leap year line for 1 y/1x = 11/30, s = 0. The jumps between steps
give the leap year placement; the dashed line shows the line being approximated.
4. COMPARING THE CALCULATIONS
Let y stand for a year number and x for a
day number; we want to assign y values
to x values, that is, determine in what
year a particular day falls. If day x falls
within year y, we darken the pixel (x, y).
The days are labeled sequentially, and no
day is in two different years, so we have
the same connectedness as Bresenham
lines. A set of days in the same year
forms a contiguous sequence of pixels at
the same y-coordinate and the lengths
of these sequences vary between two dis-
tinct lengths, L and L + 1, following the
leap year pattern. In essence, this map-
ping from days to years is a set of inte-
gral lattice points approximating a line for
which the slope is the exact ratio of years
to days. Thus we form a correspondence
of days and years with grid points and
an approximation to a line by those grid
points; the number cL + l of days in a cy-
cle is the run 1x, the number c of years
in a cycle is the rise 1 y, and the aver-
age year length L̄ is 1x/1 y, the recipro-
cal of the slope. The number of leap years
l in the run 1x is 1x mod 1 y. Hence leap
years can be viewed as discretizations of
lines with rational slopes; we must show
the Bresenham line is a special case of the
leap year calculation—we do so by show-
ing that Equation (2) is a special case of
Equation (8).
Inspecting the quotients in equations (2)
and (8), and using the fact that the exact
average year length L̄ = L +l /c is also the
reciprocal of the slope, we can rewrite (8)
Harris and Reingold
in terms of 1x and 1 y,
y = yS
¹ º
1 y(x − xS ) − ((ls − l ) mod 1 y) + 1 y − 1
+ .
1x
(9)
We call a set of (day, year) points deter-
mined by the ratio 1 y/1x and the shift s a
leap year line (whether it actually has leap
years or not). Figure 2 shows the leap year
line for 1 y/1x = 11/30, s = 0. For the
same ratio, a Bresenham line and a leap
year line have the same pattern, but they
can be shifted differently. A Bresenham
line has first and last segments close to
half the average line segment length, but a
leap year line has arbitrary end segments
depending on the shift—compare the leap
year line in Figure 2 to the Bresenham
line in Figure 1(b). When s = 0, the leap
year line spreads the two type of segments
(common years are segments of length L
and leap years are segments of length
L + 1) evenly.
To show now that Bresenham lines are
specific cases of leap year lines, we find a
leap year line whose shift s depends on the
endpoints of the Bresenham line. Assum-
ing, as we do for Bresenham lines, that the
slope is between 0 and 1, we want to solve
for s such that Equations (2) and (9) are
equal. Comparing equations, and replac-
ing l and c as needed, we find that
1x = 2(1 y − (l (s − 1) mod 1 y) − 1).
ACM Computing Surveys, Vol. 36, No. 1, March 2004.
Line Drawing, Leap Years, and Euclid 75

Table I. Summary of Comparison Between Leap Years and Bresenham Lines

(The parameters of the calendar formulas are c, L, l < c, and s; the parameters of the Bresenham line are
xS , yS , xE , and yE .)
Calendar formula/ Approximate Islamic
Calendar Bresenham line Bresenham formula calendar
Day number x-coordinate Integer x
Year number y-coordinate Integer y
Starting point xS , y S
Ending point xE , y E
Years in leap year Segments in repeating c = 1 y = yE − yE 30
cycle pattern
Days in leap year cycle Pixels in repeating cL + l = 1x = xE − xS 10631
pattern
Ratio of years to days Slope of line c/(cL + l ) = 1 y/1x 30/10631
Leap years per cycle Number of long l = 1x mod 1 y 11
segments
Days in an ordinary Pixels in short segments L = b1x/1 yc 354
year
Days in a leap year Pixels in long segments L + 1 = b1x/1 yc + 1 355
Average year length Average segment length L̄ =j(cL + l )/c = 1x/1
k y 10631/30 = 354 11
30
21 y − 1x − 1
Position of year 1 in s= +1 4
2l
cycle
Initial error 1x
Year y is a leap year yth segment is long l (jy + s) mod c < l k 11( y + 4) mod
j 30 < 11k
ly + ((l (s − 1)) mod c) 11 y + 3
Year y begins on day x Segment begins at (x, y) x = Ly + x = 354 y +
c 30
j k j k
cx − ((l (s − 1)) mod c) + c − 1 30x + 26
Day x is in year y Point (x, y) is on the line y = y=
cL + l 10631

The shift s can thus be 5. THE UNDERLYING PATTERN

$ %
21 y−1x−1
+ k1 y
s= 2
+ 1,
l the repeating pattern: 1x and 1 y (as
for any integer k. Choosing k = 0 gives
¹ º
21 y − 1x − 1
s= + 1.
2l
A summary of the comparisons between
leap year lines and Bresenham lines is
given in Table I. There are a few pecu-
liarities that distinguish Bresenham lines
and leap year lines. A Bresenham line is
defined by its two endpoints; a leap year
line by its slope (the ratio), the starting
point, and the shift. The intention of draw-
ing a Bresenham line is to plot a finite seg-
ment; the intention of the leap year line is
to give a function for an assignment for
all days, not just a range of days; thus, a
leap year pattern repeats endlessly, but a
Bresenham line does not.
For both line drawing and leap years there
are only two parameters that determine
throughout, we assume 1x ≥ 1 y). If we
are only interested in these cycles, we can
restrict ourselves to leap year lines with
no shift, that is s = 0. As mentioned
above with respect to Bresenham’s algo-
(10)
rithm, the pattern of alternating normal
and leap segments can be difficult to see
by inspection. Of course, the pattern is de-
termined (implicitly) by equation (8), but
equation (8) does not explicitly give the re-
peating pattern. The pattern is intimately
connected with Euclid’s algorithm for com-
puting the greatest common divisor of 1x
and 1 y.
Euclid’s algorithm can be expressed in
many ways, but the most basic is given
in as a function EUCLID in Algorithm 2(a),
in which we use repeated subtraction to
get the greatest common divisor. The proof
that this subtractive algorithm is correct
follows from the invariant relationship

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

76 Harris and Reingold

Algorigthm 2. Subtractive Euclid’s algorithm to compute gcd(u, v), u, v > 0, and a modified version to
compute the leap year pattern. kP kx and kP k y are, respectively, the horizontal and vertical extents of a path
P . Both algorithms take O(u + v) iterations.

Table II. A Trace of MODEUCLID (30, 11) Before the Test in Line 3 is
Executed (2 is an abbreviation for and 3 is an abbreviation
for . Figure 3 shows the resulting pattern)
u v Pl kPl k y /kPl kx Pr kPr k y /kPr kx
30 11 0/1 1/0
19 11 0/1 1/1
8 11 0/1 1/2
8 3 3 1/3 2 1/2
5 3 3 1/3 32 2/5
2 3 3 1/3 332 3/8
2 1 3332 4/11 332 3/8
1 1 3332 4/11 3332 332 7/19
return 3332 3332 332 11/30

given between lines 1 and 2, that the
greatest common divisor of u and v at that
point equals the greatest common divisor
of the original parameters to EUCLID.
We can modify the subtractive Euclid’s
algorithm, so that it builds up the leap
year pattern by concatenating appropriate
smaller patterns; the modified algorithm
is MODEUCLID in Algorithm 2(b). It com-
poses paths Pl and Pr that correspond,
respectively, to approximations from be-
low and above the desired slope. We call
MODEUCLID with the numerator and de-
nominator of that desired slope; the ini-
tial Pl = and the initial Pr = corre-
spond to the slopes 0/1 and 1/0, where
is a rightward movement of the line of
length 1 and is an upward movement. We
use the notations kP kx and kP k y as, re-
spectively, the horizontal and vertical ex-
tents of a path P ; thus
kP k y
slope(P ) = .
kP kx
An example trace of MODEUCLID is shown
in Table II. Figure 3 shows the resulting
path.
The invariant relationships given be-
tween lines 3 and 4 of MODEUCLID are
the heart of an inductive proof that

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

Line Drawing, Leap Years, and Euclid 77

Fig. 3. The output from MODEUCLID (30, 11). It is an articulated leap year line corresponding
to Figure 2. The dashed line from (0, 0) to (11, 30) touches integral lattice points only at the
endpoints because gcd(11, 30) = 1.

MODEUCLID computes the correct leap year
line. First, we prove that the relationships,
original value of u = ukPl kx + vkPr kx ,
(11)
sired. Similarly, to prove that
original value of v = u0 kPl0 k y + v0 kPr0 k y ,
we have

original value of v = ukPl k y + vkPr k y , u0 kPl0 k y +v0 kPr0 k y = ukPl Pr k y

(12) + (v − u)kPr k y
= u(kPl k y + kPr k y )
kPl kx kPr k y − kPr kx kPl k y = 1, (13) + (v − u)kPr k y
and = ukPl k y + vkPr k y ,

slope(Pl ) < slope(Pr ) (14) which, by induction, is the original value

hold from iteration to iteration; they are
clearly true at the first iteration. If u < v
in line 4, then we set v := v−u in line 5 and
Pl := Pl Pr in line 6. Let us refer to values
at the next moment we are between lines
3 and 4 by appending primes to them; thus
we must prove that
original value of u = u0 kPl0 kx + v0 kPr0 kx .
Expressing this in terms of the old values,
u0 kPl0 kx + v0 kPr0 kx = ukPl Pr kx + (v − u)kPr kx
= u(kPl kx + kPr kx )
+ (v − u)kPr kx
= ukPl kx + vkPr kx ,
which, by induction (on the number of it- (u, v) and that it is a leap year line. The
erations), is the original value of u, as de- path returned, (Pl Pr )u , goes from (0, 0) to
of v. Finally, to prove that (13) holds from
iteration to iteration,
kPl0 kx kPr0 k y −kPr0 kx kPl0 k y
= kPl Pr kx kPr k y − kPr kx kPl Pr ky
= (kPl kx +kPr kx )kPr k y − kPr kx
× (kPl k y +kPr k y ) = kPl kx kPr k y
− kPr kx kPl k y = 1,
by induction. The proofs for the case u ≥
v are almost identical. Finally, (14) holds
because (13) guarantees that
slope(Pl ) < slope(Pl Pr ) < slope(Pr ).
Now we use the invariant relation-
ships to prove that the path returned
goes from (0, 0) to the original value of

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

78 Harris and Reingold

Fig. 4. If Pl is a leap year line from (0, 0) to (kPl kx , kPl k y ), Pr is a leap year line from
(0, 0) to (kPr kx , kPr k y ), and kPl kx kPr k y − kPr kx kPl k y = 1, then the recursive structure
of the leap year line from (0, 0) to (kPl kx + kPr kx , kPl k y + kP k y ) is Pl Pr .

(ukPl kx + ukPr kx , ukPl k y + ukPr k y ), but which would mean that the leap year line
because the loop ends when u = v, from (0, 0) to (kPl Pr kx , kPl Pr k y ) passes
through the same points as Pl for 0 ≤ i ≤
(ukPl kx + ukPr kx , ukPl k y + ukPr k y ) kPl kx . Equation (15) follows from proving
= (ukPl kx + vkPr kx , ukPl k y + vkPr k y ) that if âb − ab̂ = 1 and 0 ≤ i < b, then
= (original value of u, j a k ¹ a + â º
original value of v), i = i . (16)
b b + b̂
by (11) and (12).
A path P is a leap year line if it passes Using the identity
through (i, bikP k y /kPx kc) for each i, 0 ≤
i ≤ kPx k. Initially, the paths Pl and Pr jak a a mod b
are (trivial) leap year lines. To prove that = − ,
b b b
the path returned is a leap year line, we
must show that the recursive structure proposed equation (16) becomes
shown in Figure 4 is correct—that is, that
if Pl is a leap year line from (0, 0) to ai ai mod b
(kPl kx , kPl k y ), Pr is a leap year line from −
(0, 0) to (kPr kx , kPr k y ), and kPl kx kPr k y − b b
kPr kx kPl k y = 1, then Pl Pr is a leap year (a + â)i (a + â)i mod (b + b̂)
= − ,
line from (0, 0) to (kPl kx + kPr kx , kPl k y + b + b̂ b + b̂
kP k y ).
Thus to prove that Pl Pr is a leap year or
line, we must show that it passes through
(i, bikPl Pr k y /kPl Pr kx c) for each i, 0 ≤ i ≤ (a + â)i mod (b + b̂) ai mod b
kPl Pr kx = kPl kx + kPr kx , given that Pl −
b + b̂ b
passes through (i, bikPl k y /kPl kx c) for each
i, 0 ≤ i ≤ kPl kx , and that Pr passes (a + â)i ia
= − .
through (i, bikPr k y /kPr kx c) for each i, 0 ≤ b + b̂ b
i ≤ kPr kx . First, consider the case i, 0 ≤
i ≤ kPl kx . It suffices to show that
Multiplying through by b(b + b̂) and using
¹ º ¹ º âb − ab̂ = 1 transforms this into
kPl k y kPl k y + kPr k y
i = i , (15)
kPl kx kPl kx + kPr kx
b(a + â)i mod b(b + b̂)
that is, − a(b + b̂)i mod b(b + b̂) = i.
¥ ¦ ¥ ¦
slope(Pl )i = slope(Pl Pr )i , Rearranging and again using âb − ab̂ = 1

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

Line Drawing, Leap Years, and Euclid 79

Algorigthm 3. Euclid’s algorithm to compute gcd(u, v), u, v ≥ 0, and a modified version

to compute the leap year pattern. The loop invariants (not shown) are identical to those
in Algorithm 2. By Lamé’s theorem these algorithms take O(log max(u, v)) iterations.

Table III. A Trace of MODEUCLID0 (30, 11) Before the Test in Line 5 is Executed
u v Pl kPl k y /kPl kx Pr kPr k y /kPr kx
30 11 0/1 1/0
8 11 0/1 =2 1/2
8 3 3 1/3 2 1/2
2 3 3 1/3 332 3/8
2 1 3332 4/11 332 3/8
0 1 3332 4/11 3332 3332 332 11/30
return 3332 3332 332 11/30

makes this into To prove that Pl Pr passes through

(i, bikPl Pr k y /kPl Pr kx c) for each i, kPl kx <
[a(b + b̂) + 1]i mod b(b + b̂) i ≤ kPr kx + kPr kx is similar, but uses
= i + a(b + b̂)i mod b(b + b̂), ¹ º ¹ º
â a + â
a+ (i − b) = i , (17)
which is true provided that b̂ b + b̂

i + a(b + b̂)i mod b(b + b̂) < b(b + b̂), given âb−ab̂ = 1 and b ≤ i ≤ b+ b̂. A proof
of (17) parallels our proof of (16). It follows
that is, if that MODEUCLID computes the appropriate
leap year line.
a(b + b̂)i mod b(b + b̂) < b(b + b̂) − i. Of course, we do not need to do re-
peated subtraction in Euclid’s algorithm—
Dividing this inequality by b + b̂ trans- we can rewrite it in its more common
forms it to form by using the modulus function to
group repeated subtractions into a sin-
i gle operation. Doing so yields EUCLID0 ,
ai mod b < b − ,
b + b̂ given in Algorithm 3(a). We can simi-
larly rewrite MODEUCLID; the result is
which holds because i ≤ b insures that i < MODEUCLID0 , shown as Algorithm 3(b).
b + b̂. Thus (16), and hence (15), hold. Table III given the trace of MODEUCLID0

ACM Computing Surveys, Vol. 36, No. 1, March 2004.

80
(30, 11) corresponding to that shown in
Table II.
Good practice dictates that we not give
an algorithm without discussing its run-
ning time. The subtractive EUCLID (and
hence also MODEUCLID) take time O(u + v).
From Lamé’s theorem we know that the
worst case of the gcd-form of Euclid’s algo-
rithm is a pair of adjacent Fibonacci num-
bers u = Fi+1 and v = Fi , causing i itera-
tions (see Knuth [1998], Theorem F, page
360). EUCLID0 thus takes O(log max(u, v))
iterations, as does MODEUCLID0 .
6. CONCLUSIONS
The pattern of a Bresenham line is a spe-
cial case of a leap year rule, and both
are described by Euclid’s algorithm, as
adapted in Algorithm 3. The trace of
Algorithm 3 in Table III suggests that the
slopes of the paths Pl and Pr are the con-
tinuants of the continued fraction expan-
sion of v/u; this is indeed the case. Thus
our discussion can be tied to continued
fractions and many other applications of
Euclid’s algorithm such as finding paths in
the Stern-Brocot tree or finding the short-
est factorization in elementary matrices of
a 2 × 2 integer matrix with determinant
1 (see [Graham et al. 1994, Section 6.7]).
Both line drawing and leap year calcula-
tions are essentially computing bαx + βc,
for rational α and β, and are thus also re-
lated to the Beatty sequences [Fraenkel
et al. 1978].
REFERENCES
ALLOUCHE, J.-P. AND SHALLIT, J. 2003. Automatic
Sequences. Cambridge University Press, Cam-
bridge, U.K.
Received June 2003; accepted April 2004
Harris and Reingold
BERSTEL, J. 1990. Tracé de droites, fractions con-
tinues et morphismes itérés. In Mots: Mélanges
Offerts à M.-P. Schützenberger, M. Lothaire,
Ed. Editions Hermès, Paris, France, 298–
309.
BRESENHAM, J. E. 1965. Algorithm for computer
control of a digital plotter. IBM Syst. J. 4, 1, 25–
30.
BRONS, R. 1974. Linguistic methods for the de-
scription of a straight line on a grid. Comput.
Graph. Image Process. 3, 1, 48–62.
CASTLE, C. M. A. AND PITTEWAY, M. L. V. 1987.
An efficient structural technique for encoding
‘best-fit’ straight lines. Comput. J. 30, 2, 168–
175.
FRAENKEL, A. S., MUSHKIN, M., AND TASSA, U.
1978. Determination of [nθ ] by its sequence
of differences. Can. Math. Bull. 21, 441–
446.
GRAHAM, R. L., KNUTH, D. E., AND PATASHNIK, O. 1994.
Concrete Mathematics, 2nd ed. Addison-Wesley,
Reading, MA.
KNUTH, D. E. 1998. The Art of Computer Pro-
gramming (Volume 2: Seminumerical Algo-
rithms), 3rd ed. Addison-Wesley, Reading,
MA.
PITTEWAY, M. L. V. 1985. The relationship be-
tween Euclid’s algorithm and run-length encod-
ing. In Fundamental Algorithms for Computer
Graphics, R. A. Earnshaw, Ed. Springer, Berlin,
Germany, 105–111.
REINGOLD, E. M. AND DERSHOWITZ, N. 2001. Cal-
endrical Calculations: The Millennium Edi-
tion. Cambridge University Press, Cambridge,
U.K.
ROCKETT, A. M. AND SZÜSZ, P. 1992. Continued Frac-
tions. World Scientific, Singapore.
SHALLIT, J. 1994. Pierce expansions and rules for
the determination of leap years. Fibonacci
Quart. 32, 5, 416–423.
SPROULL, R. F. 1982. Using program transforma-
tions to derive line-drawing algorithms. ACM
Trans. Graph. 1, 4, 259–273.
TROESCH, A. 1998. Droites discrètes et calendriers.
Math. Inform. et Sci. Humaines 141, 36, 11–
41.
ACM Computing Surveys, Vol. 36, No. 1, March 2004.